version: October 21, 2019
About this document
All analyses preformed with R version 3.5.0.
This is the code accompanies the publication Eckert RJ, Reaume A, Sturm AB, Studivan MS, and Voss JD.(in review) Symbiodiniaceae community variation among Montastraea cavernosa populations along a depth gradient on the Belize Barrier Reef. Here you will find all the code to repeat the statistical analyses performed for this manuscript. All of the accompanying data can be found on my github.
If you download my entire accompanying github directory you should be able to re-run these analyses by following along with the code chunks in R Studio. If you download the code separtely or you are using this pipeline on your own data, you may need to change the working directory to where the associated files are housed (ie. setwd("~/path/to/directory/with/data")).
The data used for this analysis are OTUs calculated with R packages dada2 and LULU. The raw Symbiodiniaceae ITS2 sequences obtained from Montastraea cavernosa samples can be found in the NCBI SRA under project number XXX. Hopefully you are able to follow along this file and find it useful to use with your own data!
Additional protocols and walkthroughs:
Basic setup of R environment and data
Loading required packages
For the following analyses we will require the use of a number of different R packages. Most of these can be sourced from CRAN, but a couple need to be downloaded from GitHub or BioConducter. We can use the following code to quickly load in the packages and install any packages not previously installed in the R console.
if (!require("pacman")) install.packages("pacman")
pacman::p_load("BiocManager", "car", "cluster", "dunn.test", "edgeR", "ggpubr", "labdsv",
"MCMC.OTU", "multcompView", "pgirmess", "plyr", "poppr", "reshape", "vegan")
pacman::p_load_gh("pmartinezarbizu/pairwiseAdonis/pairwiseAdonis")
if (!require("edgeR")){BiocManager::install("edgeR", update = FALSE)
library(edgeR)}Creating a color palette to use for our data
We need 10 colors, I’m using mostly ggPlot colors, but in a better order for visualizing our OTU bar chart. I like to set a default palette at the beginning and call specific colors from it as needed throughout.
its2ColPal=c("#0081EF", "#00B0F6", "#00BFC4", "#39B600", "#9590FF", "#E76BF3", "#FF62BC",
"#F8766D", "#D89000", "#A3A500")
Loading our data into R
First, we need to load in the data from dada2 and LULU analyses.
curated_result_97 = readRDS("curated_result_97.rds") #read in LULU output
its2dada <- read.csv("ASVtable.csv") #read in dada2 output
head(its2dada)Preparing data for analyses
Now that the data are loaded in, we can continue to make these data the way we need them for our analyses. We can combine our curated OTUs with the metadata we added to our ASV file.
ITS2data = cbind(its2dada[, 1:3], data.frame(t(curated_result_97$curated_table)))And we can order sites from north to south so later data can be plotted shallow to deep; north to south. To do this we can set “Depth” and “Site” as factors
ITS2data$Depth = factor(ITS2data$Depth, levels = c("10", "16", "25", "35"))
levels(ITS2data$Site)## [1] "GR" "RW" "SR" "TR"ITS2data$Site = factor(ITS2data$Site, levels(ITS2data$Site)[c(4, 2, 3, 1)])
ITS2data = ITS2data[order(ITS2data$Site, ITS2data$Depth), ]
ITS2data$newOrder = 1:nrow(ITS2data)
ITS2data = cbind(ITS2data[, length(ITS2data), drop = FALSE],
ITS2data[, c(1:length(ITS2data) - 1)])
row.names(ITS2data) = ITS2data$Sample
head(ITS2data)## newOrder Sample Site Depth sq1 sq10 sq11 sq14 sq17 sq18 sq25 sq272
## 107 1 107 TR 10 6456 0 0 0 0 0 0 0
## 108 2 108 TR 10 215412 0 0 0 0 0 0 0
## 109 3 109 TR 10 16407 0 0 0 0 0 0 0
## 110 4 110 TR 10 197896 0 0 6 3 618 1 0
## 111 5 111 TR 10 180119 0 0 0 0 0 0 0
## 112 6 112 TR 10 24462 0 0 0 0 0 0 0
## sq300 sq32 sq360 sq442 sq474 sq485 sq5 sq503 sq531 sq537 sq547 sq555
## 107 0 0 0 0 0 0 649 0 0 0 0 0
## 108 0 0 0 0 0 0 17803 0 0 0 0 0
## 109 0 0 0 0 0 0 1397 0 0 0 0 0
## 110 0 2 0 0 0 0 19950 0 0 0 0 0
## 111 0 0 0 6 0 0 16261 3 0 1 0 0
## 112 0 0 0 0 0 0 2438 0 0 0 0 0
## sq563 sq64 sq7 sq8
## 107 0 0 225 0
## 108 0 0 4736 0
## 109 0 0 266 0
## 110 0 0 7923 36
## 111 0 0 8595 0
## 112 0 0 862 0Now we can call these data for the following analyses.
Purging outlying OTUs
goods = purgeOutliers(ITS2data, count.columns = 5:length(ITS2data), otu.cut = 0.001)## [1] "samples with counts below z-score -2.5 :"
## [1] "107" "44" "212" "220" "67" "189" "186" "197"
## [1] "zscores:"
## 107 44 212 220 67 189 186
## -3.289894 -2.977162 -2.594488 -2.840026 -2.895177 -4.272415 -3.211249
## 197
## -4.032355
## [1] "OTUs passing frequency cutoff 0.001 : 10"goods$newOrder = 1:nrow(goods)
row.names(goods) = goods$sample
head(goods)## newOrder sample Site Depth sq1 sq10 sq11 sq14 sq17 sq18 sq25 sq5
## 108 1 108 TR 10 215412 0 0 0 0 0 0 17803
## 109 2 109 TR 10 16407 0 0 0 0 0 0 1397
## 110 3 110 TR 10 197896 0 0 6 3 618 1 19950
## 111 4 111 TR 10 180119 0 0 0 0 0 0 16261
## 112 5 112 TR 10 24462 0 0 0 0 0 0 2438
## 113 6 113 TR 10 195026 0 0 7 15 826 14 23224
## sq7 sq8
## 108 4736 0
## 109 266 0
## 110 7923 36
## 111 8595 0
## 112 862 0
## 113 6242 21row.names(goods) = goods$sampleSamples with data for OTUs
What is the proportion of samples with data for these OTUs?
props = apply(goods[, 5:length(goods[1, ])], 2, function(x) {
sum(x) / sum(goods[, 5:length(goods[1, ])])
})
props## sq1 sq10 sq11 sq14 sq17 sq18
## 0.848909686 0.009562495 0.007899518 0.002501895 0.001801993 0.001855771
## sq25 sq5 sq7 sq8
## 0.001136438 0.073101722 0.034023114 0.019207368props = as.data.frame(props)
barplot(props$props, log = "y", names.arg = rownames(props),
xlab = "Sequence", ylab = "Proportion", col = its2ColPal)
Normalizing reads across all samples
Here we normalize OTU counts with weighted trimmed mean of M-values (TMM; Robinson and Oshlack 2010). This helps to account for disparity in sequencing depth across libraries. First we transpose the data to work with edgeR
itsGoodsTransposed = t(goods[, 5:length(goods[1, ])])
itsGoodsList = DGEList(counts = itsGoodsTransposed)
head(itsGoodsList$samples)## group lib.size norm.factors
## 108 1 237951 1
## 109 1 18070 1
## 110 1 226433 1
## 111 1 204975 1
## 112 1 27762 1
## 113 1 225375 1Now we can use TMM normalization in edgeR
its2Norm = calcNormFactors(itsGoodsList, method = "TMM")
head(its2Norm$samples)## group lib.size norm.factors
## 108 1 237951 1.090520
## 109 1 18070 1.095378
## 110 1 226433 1.057821
## 111 1 204975 1.063382
## 112 1 27762 1.066040
## 113 1 225375 1.048051its2TMM = t(cpm(its2Norm, normalized.lib.sizes = TRUE))
its2Norm = cbind(goods[,c(2:4)], its2TMM)
head(its2Norm)## sample Site Depth sq1 sq10 sq11 sq14 sq17 sq18
## 108 108 TR 10 830134.8 0 0 0.00000 0.00000 0.000
## 109 109 TR 10 828909.5 0 0 0.00000 0.00000 0.000
## 110 110 TR 10 826199.7 0 0 25.04951 12.52476 2580.100
## 111 111 TR 10 826360.2 0 0 0.00000 0.00000 0.000
## 112 112 TR 10 826546.9 0 0 0.00000 0.00000 0.000
## 113 113 TR 10 825665.6 0 0 29.63533 63.50427 3496.969
## sq25 sq5 sq7 sq8
## 108 0.000000 68607.55 18251.16 0.00000
## 109 0.000000 70578.82 13438.77 0.00000
## 110 4.174918 83289.62 33077.88 150.29706
## 111 0.000000 74603.14 39432.63 0.00000
## 112 0.000000 82377.62 29126.13 0.00000
## 113 59.270653 98321.55 26426.24 88.90598Renaming OTU sequence columns
Finally, I want to rename the columns with the Symbiodiniaceae sequences identified through our previose BLASTn query
colnames(its2Norm)[4:ncol(its2Norm)] = c("sq01_C3", "sq10_C3", "sq11_C3g", "sq14_C3g",
"sq17_C3g", "sq18_C3g", "sq25_C3g", "sq05_C3z",
"sq07_C3e", "sq08_C3g")
head(its2Norm)## sample Site Depth sq01_C3 sq10_C3 sq11_C3g sq14_C3g sq17_C3g sq18_C3g
## 108 108 TR 10 830134.8 0 0 0.00000 0.00000 0.000
## 109 109 TR 10 828909.5 0 0 0.00000 0.00000 0.000
## 110 110 TR 10 826199.7 0 0 25.04951 12.52476 2580.100
## 111 111 TR 10 826360.2 0 0 0.00000 0.00000 0.000
## 112 112 TR 10 826546.9 0 0 0.00000 0.00000 0.000
## 113 113 TR 10 825665.6 0 0 29.63533 63.50427 3496.969
## sq25_C3g sq05_C3z sq07_C3e sq08_C3g
## 108 0.000000 68607.55 18251.16 0.00000
## 109 0.000000 70578.82 13438.77 0.00000
## 110 4.174918 83289.62 33077.88 150.29706
## 111 0.000000 74603.14 39432.63 0.00000
## 112 0.000000 82377.62 29126.13 0.00000
## 113 59.270653 98321.55 26426.24 88.90598levels(its2Norm$Site)## [1] "TR" "RW" "SR" "GR"levels(its2Norm$Depth)## [1] "10" "16" "25" "35"After all that everything looks good to go. On to the fun stuff!
OTU \(\alpha\)-diversity
Looking at \(\alpha\)-diversity of Symbiodiniaceae OTU sequences in each of our samples for comparison. We will look at species richness, Shannon’s index and Simpson’s index
Calculate \(\alpha\)-diversity metrics
its2Richness = specnumber(its2Norm[,4:ncol(its2Norm)])
its2Shannon = diversity(its2Norm[,4:ncol(its2Norm)],index = "shannon")
its2Simpson = diversity(its2Norm[,4:ncol(its2Norm)],index = "simpson")
its2Div = cbind(its2Norm[,1:3], its2Richness, its2Shannon, its2Simpson)
colnames(its2Div)[4:6] = c("richness", "shannon", "simpson")
head(its2Div)## sample Site Depth richness shannon simpson
## 108 108 TR 10 3 0.3620252 0.1744764
## 109 109 TR 10 3 0.3476677 0.1693987
## 110 110 TR 10 8 0.4670561 0.2271793
## 111 111 TR 10 3 0.4476261 0.2197705
## 112 112 TR 10 3 0.4329306 0.2149295
## 113 113 TR 10 8 0.4816535 0.2397877Statistical tests of \(\alpha\)-diversity
Now we can test whether or not there a significant differences in any of the calculated \(\alpha\)-diversity metrics across Site or Depth.
Assumption testing
First we need to see if our data are normally distributed. We can use Shapiro-Wilk’s test.
tapply(its2Div$richness, its2Div$Site, shapiro.test)## $TR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.49821, p-value = 8.713e-13
##
##
## $RW
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.59852, p-value = 2.451e-11
##
##
## $SR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.51981, p-value = 1.024e-12
##
##
## $GR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.58742, p-value = 2.111e-11tapply(its2Div$shannon, its2Div$Site, shapiro.test)## $TR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.55244, p-value = 4.961e-12
##
##
## $RW
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.61603, p-value = 4.648e-11
##
##
## $SR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.69458, p-value = 6.984e-10
##
##
## $GR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.51309, p-value = 1.8e-12tapply(its2Div$simpson, its2Div$Site, shapiro.test)## $TR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.56825, p-value = 8.468e-12
##
##
## $RW
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.63812, p-value = 1.073e-10
##
##
## $SR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.73532, p-value = 4.536e-09
##
##
## $GR
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.55991, p-value = 8.211e-12tapply(its2Div$richness, its2Div$Depth, shapiro.test)## $`10`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.7758, p-value = 3.534e-08
##
##
## $`16`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.67303, p-value = 4.331e-10
##
##
## $`25`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.2022, p-value = 3.019e-16
##
##
## $`35`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.17992, p-value = 4.725e-16tapply(its2Div$shannon, its2Div$Depth, shapiro.test)## $`10`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.67641, p-value = 3.197e-10
##
##
## $`16`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.68356, p-value = 6.73e-10
##
##
## $`25`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.71312, p-value = 1.977e-09
##
##
## $`35`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.77289, p-value = 6.555e-08tapply(its2Div$simpson, its2Div$Depth, shapiro.test)## $`10`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.70942, p-value = 1.353e-09
##
##
## $`16`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.70544, p-value = 1.734e-09
##
##
## $`25`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.74479, p-value = 8.794e-09
##
##
## $`35`
##
## Shapiro-Wilk normality test
##
## data: X[[i]]
## W = 0.83899, p-value = 2.827e-06We see that our data are NOT normally distributed.
We can use Levene’s test to test for equal variance among sites and depths.
leveneTest(its2Div$richness, its2Div$Depth)## Levene's Test for Homogeneity of Variance (center = median)
## Df F value Pr(>F)
## group 3 25.117 4.368e-14 ***
## 229
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1leveneTest(its2Div$shannon, its2Div$Depth)## Levene's Test for Homogeneity of Variance (center = median)
## Df F value Pr(>F)
## group 3 6.7836 0.0002115 ***
## 229
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1leveneTest(its2Div$simpson, its2Div$Depth)## Levene's Test for Homogeneity of Variance (center = median)
## Df F value Pr(>F)
## group 3 6.6744 0.0002442 ***
## 229
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1leveneTest(its2Div$richness, its2Div$Site)## Levene's Test for Homogeneity of Variance (center = median)
## Df F value Pr(>F)
## group 3 1.0165 0.3861
## 229leveneTest(its2Div$shannon, its2Div$Site) ## Levene's Test for Homogeneity of Variance (center = median)
## Df F value Pr(>F)
## group 3 0.4804 0.6962
## 229leveneTest(its2Div$simpson, its2Div$Site)## Levene's Test for Homogeneity of Variance (center = median)
## Df F value Pr(>F)
## group 3 0.5514 0.6477
## 229Data are non-normally distributed and also heteroschedastic. Using non-parametric tests moving forward.
Kruskal-Wallis Tests
Since our data did not fit the assumptions of ANOVA, we will use single factor Krskal-Wallis tests on these data.
K-W test on Shannon’s indices
kruskal.test(shannon ~ Site, data = its2Div)##
## Kruskal-Wallis rank sum test
##
## data: shannon by Site
## Kruskal-Wallis chi-squared = 2.7884, df = 3, p-value = 0.4254kruskal.test(shannon ~ Depth, data = its2Div)##
## Kruskal-Wallis rank sum test
##
## data: shannon by Depth
## Kruskal-Wallis chi-squared = 3.0743, df = 3, p-value = 0.3803Nothing significant
K-W test on Simpson’s indices
kruskal.test(simpson ~ Site, data = its2Div)##
## Kruskal-Wallis rank sum test
##
## data: simpson by Site
## Kruskal-Wallis chi-squared = 3.4937, df = 3, p-value = 0.3216kruskal.test(simpson ~ Depth, data = its2Div)##
## Kruskal-Wallis rank sum test
##
## data: simpson by Depth
## Kruskal-Wallis chi-squared = 3.8446, df = 3, p-value = 0.2787Again, no significant tests.
K-W test on species richness
kruskal.test(richness ~ Site, data = its2Div)##
## Kruskal-Wallis rank sum test
##
## data: richness by Site
## Kruskal-Wallis chi-squared = 2.7628, df = 3, p-value = 0.4297kruskal.test(richness ~ Depth, data = its2Div)##
## Kruskal-Wallis rank sum test
##
## data: richness by Depth
## Kruskal-Wallis chi-squared = 38.993, df = 3, p-value = 1.741e-08Here we see that species richness changes significantly with Depth.
Pairwise comparisons with Dunn’s test
We now use Dunn’s test to do multiple comparisons with bonferroni correction. This will reveal where the differences in species richness are across Depth
dunn.test(its2Div$richness, g = its2Div$Depth, method = "bonferroni", table = FALSE,
list = TRUE)## Kruskal-Wallis rank sum test
##
## data: x and group
## Kruskal-Wallis chi-squared = 38.9935, df = 3, p-value = 0
##
##
## Comparison of x by group
## (Bonferroni)
##
## List of pairwise comparisons: Z statistic (adjusted p-value)
## -----------------------------
## 10 - 16 : 1.449108 (0.4419)
## 10 - 25 : 4.989355 (0.0000)*
## 16 - 25 : 3.504135 (0.0014)*
## 10 - 35 : 5.057011 (0.0000)*
## 16 - 35 : 3.591117 (0.0010)*
## 25 - 35 : 0.133160 (1.0000)
##
## alpha = 0.05
## Reject Ho if p <= alpha/2This shows us that species richness is different between our deepest sites and our shallowest sites.
Plotting \(\alpha\)-diversity metrics
We can take all of our \(\alpha\)-diversity metrics and plot them all into a single figure panel to visualize the results of the previous tests. Here we make plots for richness, Simpson’s and Shannon’s indices by site and depth zone.
Plot species richness by site
its2RichPlotS = ggplot(data = its2Div, aes(x = Site, y = richness)) +
geom_point(aes(fill = Depth),shape = 21, color = "gray20", size = 2,
position = position_jitterdodge()) +
geom_boxplot(alpha = 0, outlier.shape = NA, color = "gray30") +
scale_fill_manual(values = its2ColPal[c(1, 4, 6, 2)],
labels = c("10 m", "16 m", "25 m", "35 m")) +
xlab("Reef Site") +
ylab("Richness") +
stat_compare_means(size = 5, geom = "label", label.x = 1, label.y = 10.7) +
coord_cartesian(ylim = c(1, 11)) +
scale_y_continuous(breaks = seq(2, 10, 2)) +
guides(fill = guide_legend(ncol = 2)) +
theme_bw()
its2RichPlotSite = its2RichPlotS +
theme(axis.title.x = element_blank(),
axis.text.x = element_blank(),
axis.title.y = element_text(color = "black", size = 14),
axis.text.y = element_text(color = "black", size = 12),
legend.title = element_text(color = "black", size = 14),
legend.text = element_text(color = "black", size = 12),
panel.border = element_rect(size = 1.1, color = "black"),
panel.background = element_rect(fill = "white"),
plot.background = element_rect(fill = "white"),
legend.position = "bottom",
legend.background = element_rect(color = "black"),
plot.margin = unit(c(0.5, 0.1, 0.1, 0.61), "cm")
)
its2RichPlotSiteCreate pairwise comparison letters
We know from our Kruskal-Wallis tests that there was a significant difference in the species diversity across depth zones. We can plot letters to denote similarity between groups on our plot.
richDkmc = kruskalmc(its2Div$richness ~ its2Div$Depth) # multiple-comparison test
print(richDkmc)## Multiple comparison test after Kruskal-Wallis
## p.value: 0.05
## Comparisons
## obs.dif critical.dif difference
## 10-16 14.040805 32.74642 FALSE
## 10-25 48.134463 32.60500 TRUE
## 10-35 49.441667 33.04241 TRUE
## 16-25 34.093659 32.88254 TRUE
## 16-35 35.400862 33.31630 TRUE
## 25-35 1.307203 33.17732 FALSErichDkmcDiff = richDkmc$dif.com$difference # select logical vector
names(richDkmcDiff) = row.names(richDkmc$dif.com)# add comparison names
# create a list with "homogenous groups" coded by letter
richDsigLetters = multcompLetters(richDkmcDiff, compare="<", threshold=0.05,
Letters=c(letters, LETTERS, "."), reversed = FALSE)
richLetters = as.data.frame(richDsigLetters$Letters)
richLetters$depth = row.names(richLetters)
colnames(richLetters)[1] = "id"
richLetters$y = c(1, 1, 1, 1)
head(richLetters)## id depth y
## 10 a 10 1
## 16 a 16 1
## 25 b 25 1
## 35 b 35 1These letters can now be added to the corresponding plot.
Plot species richness by depth
its2RichPlotD = ggplot(data = its2Div, aes(x = Depth, y = richness)) +
geom_point(aes(fill = Site), shape = 21, color = "gray 20", size = 2,
position = position_jitterdodge()) +
geom_boxplot(alpha = 0, outlier.shape = NA, color = "gray30") +
scale_fill_manual(values = its2ColPal[c(8,5,3,9)],
labels = c("Tobacco Reef", "Raph's Wall",
"South Reef", "Gover's Reef")) +
stat_compare_means(size = 5, geom = "label", label.x = 1.07,
label.y = 10.7) +
geom_text(data = richLetters, aes(x = depth, y = y, label = id),
size = 6) +
coord_cartesian( ylim = c(1, 11)) +
scale_y_continuous(breaks = seq(2, 10, 2))+
xlab("Depth (m)") +
ylab("Richness") +
guides(fill=guide_legend(ncol=2))+
theme_bw()
its2RichPlotDepth = its2RichPlotD +
theme(axis.title.x = element_blank(),
axis.text.x = element_blank(),
axis.title.y = element_blank(),
axis.text.y = element_blank(),
legend.title = element_text(color = "black", size = 14),
legend.text = element_text(color = "black", size = 12),
panel.border = element_rect(size = 1.1, color = "black"),
panel.background = element_rect(fill = "white"),
plot.background = element_rect(fill = "white"),
legend.position = "bottom",
legend.background = element_rect(color = "black"),
plot.margin = unit(c(0.5, 0.5, 0.1, 0.85), "cm")
)
its2RichPlotDepthMake legends grobs
Now we want to take the Site and Depth legends and make them graphical objects (grobs). This will allow us to place them separately in our figure grid, so each plot won’t have its own legend.
get_legend <- function(its2RichPlotSite) {
tmp <- ggplot_gtable(ggplot_build(its2RichPlotSite))
leg <-
which(sapply(tmp$grobs, function(x)
x$name) == "guide-box")
legend <- tmp$grobs[[leg]]
return(legend)
}
legend.site = get_legend(its2RichPlotSite)
get_legend = function(its2RichPlotDepth) {
tmp <- ggplot_gtable(ggplot_build(its2RichPlotDepth))
leg <-
which(sapply(tmp$grobs, function(x)
x$name) == "guide-box")
legend <- tmp$grobs[[leg]]
return(legend)
}
legend.depth = get_legend(its2RichPlotDepth)
# re-plot without legends
its2RichPlotSite = its2RichPlotSite + theme(legend.position = "none")
its2RichPlotDepth = its2RichPlotDepth + theme(legend.position = "none")Plot Shannon’s index by site
its2ShannonPlotS = ggplot(data = its2Div, aes(x = Site, y = shannon)) +
geom_point(aes(fill = Depth), shape = 21, color = "gray20",
size = 2, position = position_jitterdodge()) +
geom_boxplot(alpha = 0, outlier.shape = NA, color = "gray30") +
scale_fill_manual(values = its2ColPal[c(1, 4, 6, 2)],
labels = c("10 m", "16 m", "25 m", "35 m")) +
stat_compare_means(size = 5, geom = "label", label.x = 1,
label.y = 1.7) +
expand_limits(y = c(0, 1.75)) +
xlab("Reef Site") +
ylab("Shannon's index") +
theme_bw()
its2ShannonPlotSite = its2ShannonPlotS +
theme(axis.title.x = element_blank(),
axis.text.x = element_blank(),
axis.title.y = element_text(color = "black", size = 14),
axis.text.y = element_text(color = "black", size = 12),
legend.title = element_text(color = "black", size = 14),
legend.text = element_text(color = "black", size = 12),
panel.border = element_rect(size = 1.1, color = "black"),
panel.background = element_rect(fill = "white"),
plot.background = element_rect(fill = "white"),
legend.position = "none",
legend.background = element_rect(color = "black"),
plot.margin = unit(c(0.1, 0.1, 0.1, 0.5), "cm")
)
its2ShannonPlotSitePlot Shannon’s index by depth
its2ShannonPlotD = ggplot(data = its2Div, aes(x = Depth, y = shannon))+
geom_point(aes(fill = Site), shape = 21, color = "gray20", size = 2,
position = position_jitterdodge()) +
geom_boxplot(alpha = 0, outlier.shape = NA, color = "gray30") +
scale_fill_manual(values = its2ColPal[c(8,5,3,9)],
labels = c("Tobacco Reef", "Raph's Wall",
"South Reef", "Gover's Reef")) +
stat_compare_means(size = 5, geom = "label", label.x = 1,
label.y = 1.7) +
expand_limits(y = c(0, 1.75)) +
xlab("Depth (m)") +
ylab("Shannon's index") +
theme_bw()
its2ShannonPlotDepth = its2ShannonPlotD +
theme(axis.title.x = element_blank(),
axis.text.x = element_blank(),
axis.title.y = element_blank(),
axis.text.y = element_blank(),
legend.title = element_text(color = "black", size = 14),
legend.text = element_text(color = "black", size = 12),
panel.border = element_rect(size = 1.1, color = "black"),
panel.background = element_rect(fill = "white"),
plot.background = element_rect(fill = "white"),
legend.position = "none",
legend.background = element_rect(color = "black"),
plot.margin = unit(c(0.1, 0.5, 0.1, 0.85), "cm")
)
its2ShannonPlotDepthPlot Simpson’s index by site
its2SimpsonPlotS = ggplot(data = its2Div, aes(x = Site, y = simpson)) +
geom_point(aes(fill = Depth), shape = 21, color = "gray20", size = 2,
position = position_jitterdodge()) +
geom_boxplot(alpha = 0, outlier.shape = NA) +
scale_fill_manual(values = its2ColPal[c(1,4,6,2)],
labels = c("10 m", "16 m", "25 m", "35 m")) +
stat_compare_means(size = 5, geom = "label", label.x = 1,
label.y = 0.77) +
expand_limits(y = c(0, 0.8)) +
xlab("Reef Site") +
ylab("Simpson's index") +
theme_bw()
its2SimpsonPlotSite = its2SimpsonPlotS +
theme(axis.title.x = element_text(color = "black", size = 14),
axis.text.x = element_text(color = "black", size = 12),
axis.title.y = element_text(color = "black", size = 14),
axis.text.y = element_text(color = "black", size = 12),
legend.title = element_text(color = "black", size = 14),
legend.text = element_text(color = "black", size = 12),
panel.border = element_rect(size = 1.1, color = "black"),
panel.background = element_rect(fill = "white"),
plot.background = element_rect(fill = "white"),
legend.position = "none",
legend.background = element_rect(color = "black"),
plot.margin = unit(c(0.11, 0.1, 0, 0.5), "cm")
)
its2SimpsonPlotSitePlot Simpson’s index by depth
its2SimpsonPlotD = ggplot(data = its2Div, aes(x = Depth, y = simpson)) +
geom_point(aes(fill = Site), shape = 21, color = "gray20", size = 2,
position = position_jitterdodge()) +
geom_boxplot(alpha = 0, outlier.shape = NA) +
scale_fill_manual(values = its2ColPal[c(8,5,3,9)],
labels = c("Tobacco Reef", "Raph's Wall",
"South Reef", "Gover's Reef")) +
stat_compare_means(size = 5, geom = "label", label.x = 1,
label.y = 0.77) +
expand_limits(y = c(0, 0.8)) +
xlab("Depth (m)") +
ylab("Simpson's index") +
theme_bw()
its2SimpsonPlotDepth = its2SimpsonPlotD +
theme(axis.title.x = element_text(color = "black", size = 14),
axis.text.x = element_text(color = "black", size = 12),
axis.title.y = element_blank(),
axis.text.y = element_blank(),
legend.title = element_text(color = "black", size = 14),
legend.text = element_text(color = "black", size = 12),
panel.border = element_rect(size = 1.1, color = "black"),
panel.background = element_rect(fill = "white"),
plot.background = element_rect(fill = "white"),
legend.position = "none",
legend.background = element_rect(color = "black"),
plot.margin = unit(c(0.1, 0.5, 0, 0.85), "cm")
)
its2SimpsonPlotDepthCreate a single figure panel
Now we want to throw all those bad boys into a single figure grid.
divPlots = ggarrange(its2RichPlotSite, its2RichPlotDepth, its2ShannonPlotSite,
its2ShannonPlotDepth, its2SimpsonPlotSite, its2SimpsonPlotDepth,
legend.site, legend.depth, ncol = 2, nrow = 4, widths = c(3.5,3.4),
heights = c(3,3,3.4,1),
labels = c("a","b", "c", "d", "e", "f","",""),
font.label = list(size = 18, color = "black", face ="bold")
)
divPlotsSave as high-res images
Finally, we will save the resulting \(\alpha\)-diversity plot panel as a high resoultion .eps figure. Alternatively you could save it as a HQ .tiff by simply changing the extension in the code below.
ggsave("its2_diversityPlots.eps", plot = divPlots, width = 8.25, height = 10,
unit = "in", dpi = 600)
Ordination with PCoA
We can use principal coordinates analysis (PCoA) to visualize our samples based on significant OTUs. But, first we need to set up data for PCoA analysis in R.
Create distance matrix
We will create a distance matrix with Bray-Curtis similarity using the package vegan
its2Dist = vegdist(its2Norm[, c(4:ncol(its2Norm))], method = "bray")Perform PCoA
This is the actual PCoA step
its2Mds = cmdscale(its2Dist, eig = TRUE, x.ret = TRUE)Determine percent variation captured on each axis
Calculate the eigenvalues so later we can figure out % variation shown on each Principal Coordinate
its2Var = round(its2Mds$eig/sum(its2Mds$eig)*100, 1)
its2Var## [1] 83.3 10.5 3.0 1.8 1.3 0.9 0.4 0.3 0.2 0.2 0.1 0.1 0.1 0.0
## [15] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [29] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [43] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [57] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [71] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [85] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [99] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [113] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [127] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [141] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [155] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [169] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [183] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [197] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [211] 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
## [225] 0.0 0.0 -0.1 -0.1 -0.1 -0.2 -0.3 -0.4 -0.6Format data to plot
its2Values = its2Mds$points
its2Values =as.data.frame(its2Values, sample = rownames(its2Values))
its2Pcoa = data.frame(sample = rownames(its2Values), site = as.factor(its2Norm$Site),
depth = as.factor(its2Norm$Depth),PCo1 = its2Values[,1],
PCo2 = its2Values[,2])
head(its2Pcoa)## sample site depth PCo1 PCo2
## 1 108 TR 10 -0.04147518 0.004611102
## 2 109 TR 10 -0.04211077 0.005071409
## 3 110 TR 10 -0.03039080 -0.004883738
## 4 111 TR 10 -0.03416250 -0.004766676
## 5 112 TR 10 -0.03216465 -0.003956257
## 6 113 TR 10 -0.02452031 -0.006515757PCoA biplot
its2PcoaA = ggplot(its2Pcoa, aes(x = PCo1, y = PCo2)) +
geom_hline(yintercept = 0, color = "gray90", size = 0.5) +
geom_vline(xintercept = 0, color = "gray90", size = 0.5) +
geom_point(aes(shape = factor(depth), size = depth, fill = site),
color = "black") +
scale_fill_manual(values = its2ColPal[c(8,5,3,9)], name = "Reef Site",
labels = c("Tobacco Reef", "Raph's Wall", "South Reef",
"Glover's Reef")) +
scale_shape_manual(values = c(24,23,22,21), name = "Depth",
labels = c("10 m", "16 m", "25 m", "35 m")) +
scale_size_manual(values = c(4, 4.75, 4.75, 4.75)) +
guides(shape = guide_legend(override.aes = list(size = c(4, 4.75, 4.75, 4.75),
fill = "white")), fill = guide_legend(override.aes =
list(shape = 22, size = 5.75, color = "white")), size = FALSE,
color = FALSE)+
xlab(paste ("PCo 1 (", its2Var[1],"%)", sep = "")) +
ylab(paste ("PCo 2 (", its2Var[2],"%)", sep = "")) +
theme_bw()
its2Pcoa = its2PcoaA +
theme(axis.title.x = element_text(color = "black", size = 12),
axis.text.x = element_blank(),
axis.ticks.x = element_blank(),
axis.line.x = element_blank(),
axis.title.y = element_text(color = "black", size = 12),
axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
axis.line.y = element_blank(),
legend.position = "right",
legend.title = element_text(color = "black", size = 12),
legend.text = element_text(color = "black", size = 12),
legend.key = element_blank(),
panel.border = element_rect(color = "black", size = 1.2),
panel.background = element_rect(fill = "white"),
plot.background = element_rect(fill = "white"),
panel.grid.major = element_blank(),
panel.grid.minor = element_blank()
)
its2PcoaASave the generated PCoA plot
ggsave("ITS2_OTU_PCoA.eps", plot = its2Pcoa, width = 10, height = 6.25, dpi = 600,
device="eps")
Ordination with nMDS
Now we can view the data with nMDS using the package vegan again. First we have to run the calculations for the nMDS. This is similar to PCoA, but can be scaled freely and uses an iterative process. We’ll run it with 50 iterations, which should be enough to arrive at a solution.
set.seed(694)
its2Nmds = metaMDS(its2Norm[4:ncol(its2Norm)], try = 50)
its2NmdsPrepare data to plot
its2Scores = as.data.frame(scores(its2Nmds))
its2Scores$site = factor(its2Norm$Site)
its2Scores$depth = as.factor(its2Norm$Depth)
its2Scores$sample = row.names(its2Scores)
head(its2Scores)## NMDS1 NMDS2 site depth sample
## 108 -0.052182197 -0.058288700 TR 10 108
## 109 -0.047954285 -0.091459268 TR 10 109
## 110 -0.023126989 0.009458743 TR 10 110
## 111 -0.069960350 0.011250979 TR 10 111
## 112 -0.056921489 -0.016495432 TR 10 112
## 113 0.002605697 -0.011187856 TR 10 113its2Clades = as.data.frame(scores(its2Nmds,"species"))
its2Clades$seq = row.names(its2Clades)
its2Clades## NMDS1 NMDS2 seq
## sq01_C3 -0.074686575 -0.03013410 sq01_C3
## sq10_C3 0.030677068 0.73232380 sq10_C3
## sq11_C3g 1.737397733 0.77028008 sq11_C3g
## sq14_C3g 1.528572689 -0.64806195 sq14_C3g
## sq17_C3g 1.582265406 0.10068409 sq17_C3g
## sq18_C3g 0.511133089 -0.33314875 sq18_C3g
## sq25_C3g 1.610640952 -0.01028774 sq25_C3g
## sq05_C3z -0.006264349 -0.04142266 sq05_C3z
## sq07_C3e -0.049458261 0.03970903 sq07_C3e
## sq08_C3g 1.670959017 -0.12912196 sq08_C3gits2Clades$seq## [1] "sq01_C3" "sq10_C3" "sq11_C3g" "sq14_C3g" "sq17_C3g" "sq18_C3g"
## [7] "sq25_C3g" "sq05_C3z" "sq07_C3e" "sq08_C3g"Construct nMDS biplot
its2NmdsPlotA = ggplot() +
geom_point(data = its2Scores, aes(x = NMDS1, y = NMDS2,
shape = depth, size = depth, fill = site),
color = "black") + # add the site points
scale_fill_manual(values = its2ColPal[c(8,5,3,9)],
name = "Reef Site", labels = c("Tobacco Reef",
"Raph's Wall", "South Reef", "Glover's Reef")) +
scale_shape_manual(values = c(24, 23, 22, 21), name = "Depth",
labels = c("10 m", "16 m", "25 m", "35 m")) +
scale_size_manual(values = c(3, 3.75, 3.75, 3.75)) +
guides(shape = guide_legend( override.aes = list(size =
c(3, 3.75, 3.75, 3.75), fill = "white")), fill =
guide_legend(override.aes = list(shape = 22, size = 3.75,
color = NA)), size = FALSE)+
geom_text(data = its2Clades, aes(x = NMDS1, y = NMDS2,
label = seq),
color = "#000000", size = 4, fontface = "italic") +
# add seq labels
annotate("label", x = 1.25, y = 0.725, label = paste
("Distance = Bray-Curtis\nStress = ",
round(its2Nmds$stress, 4), sep = ""), size = 4) +
labs(x = "nMDS1", y = "nMDS2") +
coord_equal() +
theme_bw()
its2NmdsPlot = its2NmdsPlotA +
theme(axis.title.x = element_text(color = "black", size = 12),
axis.text.x = element_blank(),
axis.ticks.x = element_blank(),
axis.title.y = element_text(color = "black", size = 12),
axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
legend.position = "right",
legend.title = element_text(color = "black", size = 12),
legend.text = element_text(color = "black", size = 12),
legend.key = element_blank(),
legend.background = element_blank(),
panel.border = element_rect(color = "black", size = 1.2),
panel.background = element_rect(fill = "white"),
plot.background = element_blank()
)
its2NmdsPlotSaving nMDS plot
ggsave("nMDS_ITS2_OTU.eps", plot = its2NmdsPlot, width = 8, height = 5, dpi = 600,
device = "eps")
Remove outlying samples
Samples with the C3g types skewing the data: 42, 80, 104, 98, 20, 57, 149, 152, 58, 163 let’s look at these samples specifically
itsOuts = its2Norm[its2Norm$sample %in% c("42", "80", "104", "98", "20", "57",
"149", "152", "58", "163"), ]
itsOuts$sums = apply(itsOuts[,4:13], 1, function(x) sum(x))
itsNoOut = its2Norm
itsNoOut <- itsNoOut[!itsNoOut$sample %in% c("42", "80", "104", "98", "20", "57",
"149", "152", "58", "163"), ]
itsNoOut$sums <- apply(itsNoOut[,4:13], 1, function(x) sum(x))
itsOuts## sample Site Depth sq01_C3 sq10_C3 sq11_C3g sq14_C3g sq17_C3g
## 98 98 TR 16 5962.282 2442.380 1152540 0.0 107105.56
## 104 104 TR 16 14559.120 3065.078 1047363 0.0 306539.72
## 42 42 RW 10 3327.416 3104.598 1241319 0.0 200848.17
## 20 20 RW 16 34844.366 0.000 0 0.0 219056.79
## 80 80 SR 10 7562.987 9943.468 2905229 0.0 187462.89
## 57 57 SR 16 59249.888 0.000 0 438911.4 361382.30
## 58 58 SR 16 69160.226 0.000 0 704896.7 155133.96
## 152 152 GR 10 6045.766 0.000 0 560943.6 79987.07
## 163 163 GR 10 171477.140 0.000 0 1164666.6 101868.88
## 149 149 GR 16 14587.886 0.000 0 499950.3 168854.14
## sq18_C3g sq25_C3g sq05_C3z sq07_C3e sq08_C3g sums
## 98 0.00 61370.79 169218.7 13145.75 884093.8 2395879
## 104 0.00 143260.47 165801.6 61908.19 1215750.4 2958247
## 42 0.00 129932.62 130690.2 38369.26 596691.8 2344283
## 20 0.00 122446.05 383743.7 22680.39 5925715.7 6708487
## 80 0.00 107791.16 164873.1 31417.39 947977.0 4362257
## 57 184315.48 189620.65 346779.9 47641.53 2467484.7 4095386
## 58 312760.62 116130.53 474786.2 20674.75 4183582.7 6037126
## 152 62068.54 53954.48 187677.3 15253.63 1879099.6 2845030
## 163 158405.47 281910.22 621065.3 0.00 5294413.0 7793807
## 149 169574.53 92441.40 243388.7 29973.35 1338843.8 2557614max(itsOuts$sum)## [1] 7793807max(itsNoOut$sum)## [1] 1372236mean(itsNoOut$sum)## [1] 942949summary(itsOuts$sum)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 2344283 2629468 3526817 4209812 5618408 7793807summary(itsNoOut$sum)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 871864 923180 932594 942949 943096 1372236Nothing unusual here, neither the highest nor the lowest sequencing depth of all samples
nMDS without outlying samples
Plotting nMDS the same way as above, just with the new data lacking 10 C3g dominated samples. This way we can see if these samples are skewing our plot and obscuring a larger pattern in rest of the data.
set.seed(694)
its2Nmds2 = metaMDS(itsNoOut[4:ncol(itsNoOut)], try = 50)
its2Nmds2
its2Scores2 = as.data.frame(scores(its2Nmds2))
its2Scores2$site = factor(itsNoOut$Site)
its2Scores2$depth = as.factor(itsNoOut$Depth)
its2Scores2$sample = row.names(its2Scores2)
head(its2Scores2)
its2Clades2 = as.data.frame(scores(its2Nmds2,"species"))
its2Clades2$seq = row.names(its2Clades2)
its2Clades2
its2Clades2$seq
## Plot the data
its2NmdsPlotB = ggplot() +
geom_point(data = its2Scores2, aes(x = NMDS1, y = NMDS2,
shape = depth, size = depth, fill = site),
color = "black") + # add the site points
scale_fill_manual(values = its2ColPal[c(8,5,3,9)], name =
"Reef Site",labels = c("Tobacco Reef",
"Raph's Wall", "South Reef", "Glover's Reef")) +
scale_shape_manual(values = c(24, 23, 22, 21), name = "Depth",
labels = c("10 m", "16 m", "25 m", "35 m")) +
scale_size_manual(values = c(3, 3.75, 3.75, 3.75)) +
guides(shape = guide_legend( override.aes = list(size =
c(3, 3.75, 3.75, 3.75), fill = "white")), fill =
guide_legend(override.aes = list(shape = 22, size = 3.75,
color = NA)), size = FALSE)+
geom_text(data = its2Clades2, aes(x = NMDS1, y = NMDS2,
label = seq), color = "#000000", size = 4,
fontface = "italic") +
# add seq labels
annotate("label", x = 0.95, y = 1.17, label = paste
("Distance = Bray-Curtis\nStress = ",
round(its2Nmds2$stress, 4), sep = ""), size = 4) +
labs(x = "nMDS1", y = "nMDS2") +
coord_equal() +
theme_bw()
its2NmdsPlot2 = its2NmdsPlotB +
theme(axis.title.x = element_text(color = "black", size = 12),
axis.text.x = element_blank(),
axis.ticks.x = element_blank(),
axis.title.y = element_text(color = "black", size = 12),
axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
legend.position = "right",
legend.title = element_text(color = "black", size = 12),
legend.text = element_text(color = "black", size = 12),
legend.key = element_blank(),
legend.background = element_blank(),
panel.border = element_rect(color = "black", size = 1.2),
panel.background = element_rect(fill = "white"),
plot.background = element_blank()
)
its2NmdsPlot2# saving new nMDS plot
ggsave("nMDS_OTU_noOut.eps", plot = its2NmdsPlot2, width = 9.2, height = 5.75,
dpi = 600, device = "eps")
Here we can see that there is not a larger pattern in the data that was being masked by the C3g dominant samples
nMDS with all original ASVs
ASVs were clustered at 97% to produce OTUs for these analyses. While this is biologically relevant for Symbiodiniaceae, we can still look at the original ASVs to see if clustering masked any patterns in the data.
Prep the data
its2Asv = its2dada
its2AsvTransposed = t(its2Asv[, 4:length(its2Asv[1, ])])
its2AsvList = DGEList(counts = its2AsvTransposed)
head(its2AsvList$samples)## group lib.size norm.factors
## Sample1 1 38585 1
## Sample2 1 198292 1
## Sample3 1 139636 1
## Sample4 1 269753 1
## Sample5 1 180421 1
## Sample6 1 187055 1Normalization in edgeR
its2AsvNorm = calcNormFactors(its2AsvList, method = "TMM")
head(its2AsvNorm$samples)## group lib.size norm.factors
## Sample1 1 38585 1.2512419
## Sample2 1 198292 0.8950909
## Sample3 1 139636 1.2843977
## Sample4 1 269753 1.2945265
## Sample5 1 180421 1.2053320
## Sample6 1 187055 1.1701927its2AsvTMM = t(cpm(its2AsvNorm, normalized.lib.sizes = TRUE))
its2AsvNorm = cbind(its2Asv[,c(1:3)], its2AsvTMM)
its2AsvNorm$Depth = factor(its2AsvNorm$Depth, levels = c("10", "16", "25", "35"))
levels(its2AsvNorm$Depth)## [1] "10" "16" "25" "35"its2AsvNorm$Site = factor(its2AsvNorm$Site, levels(its2AsvNorm$Site)[c(4, 2, 3, 1)])
levels(its2AsvNorm$Site)## [1] "TR" "RW" "SR" "GR"set.seed(694)
its2AsvNmds = metaMDS(its2AsvNorm[4:ncol(its2AsvNorm)], try = 200)its2AsvNmds##
## Call:
## metaMDS(comm = its2AsvNorm[4:ncol(its2AsvNorm)], try = 200)
##
## global Multidimensional Scaling using monoMDS
##
## Data: wisconsin(sqrt(its2AsvNorm[4:ncol(its2AsvNorm)]))
## Distance: bray
##
## Dimensions: 2
## Stress: 0.1665978
## Stress type 1, weak ties
## Two convergent solutions found after 200 tries
## Scaling: centring, PC rotation, halfchange scaling
## Species: expanded scores based on 'wisconsin(sqrt(its2AsvNorm[4:ncol(its2AsvNorm)]))'its2AsvScores = as.data.frame(scores(its2AsvNmds))
its2AsvScores$site = factor(its2AsvNorm$Site)
its2AsvScores$depth = as.factor(its2AsvNorm$Depth)
its2AsvScores$sample = row.names(its2AsvScores)
head(its2AsvScores)## NMDS1 NMDS2 site depth sample
## Sample1 -0.07210627 0.35712964 RW 25 Sample1
## Sample2 0.21816644 -0.71571178 RW 25 Sample2
## Sample3 -0.12838807 0.27708728 RW 25 Sample3
## Sample4 -0.14233900 0.26758410 RW 25 Sample4
## Sample5 -0.22086866 0.29531065 RW 25 Sample5
## Sample6 -0.65961404 0.03035301 RW 25 Sample6its2AsvClades = as.data.frame(scores(its2AsvNmds,"species"))
its2AsvClades$seq = row.names(its2AsvClades)
head(its2AsvClades)## NMDS1 NMDS2 seq
## sq1 -0.09121343 0.1495663 sq1
## sq2 -0.15315572 0.2187199 sq2
## sq3 -0.13334276 0.1968000 sq3
## sq4 -0.17039672 0.1772061 sq4
## sq5 0.03756152 0.2092348 sq5
## sq6 -0.17731745 0.1520387 sq6head(its2AsvClades$seq)## [1] "sq1" "sq2" "sq3" "sq4" "sq5" "sq6"its2AsvNmdsPlotA = ggplot() +
geom_point(data = its2AsvScores, aes(x = NMDS1, y = NMDS2,
shape = depth, size = depth, fill = site),
color = "black") + # add the site points
scale_fill_manual(values = its2ColPal[c(8,5,3,9)],
name = "Reef Site", labels = c("Tobacco Reef",
"Raph's Wall", "South Reef", "Glover's Reef")) +
scale_shape_manual(values = c(24, 23, 22, 21), name = "Depth",
labels = c("10 m", "16 m", "25 m", "35 m")) +
scale_size_manual(values = c(3, 3.75, 3.75, 3.75)) +
guides(shape = guide_legend( override.aes = list(size =
c(3, 3.75, 3.75, 3.75), fill = "white")), fill =
guide_legend(override.aes = list(shape = 22, size = 3.75,
color = NA)), size = FALSE)+
#geom_text(data = its2AsvClades, aes(x = NMDS1, y = NMDS2,
# label = seq),
# color = "#000000", size = 4, fontface = "italic") +
# turning off seq labels, they completely obscure the plot
annotate("label", x = 3.15, y = -2, label = paste
("Distance = Bray-Curtis\nStress = ",
round(its2AsvNmds$stress, 4), sep = ""), size = 4) +
labs(x = "nMDS1", y = "nMDS2") +
coord_equal() +
theme_bw()
its2AsvNmdsPlot = its2AsvNmdsPlotA +
theme(axis.title.x = element_text(color = "black", size = 12),
axis.text.x = element_blank(),
axis.ticks.x = element_blank(),
axis.title.y = element_text(color = "black", size = 12),
axis.text.y = element_blank(),
axis.ticks.y = element_blank(),
legend.position = "right",
legend.title = element_text(color = "black", size = 12),
legend.text = element_text(color = "black", size = 12),
legend.key = element_blank(),
legend.background = element_blank(),
panel.border = element_rect(color = "black", size = 1.2),
panel.background = element_rect(fill = "white"),
plot.background = element_blank()
)
its2AsvNmdsPlotggsave("nMDS_ITS2_ASV.eps", plot = its2AsvNmdsPlot, width = 8, height = 5, dpi = 600,
device = "eps")
Even with all unfiltered ASVs we still see the same clustering pattern among all samples.
Relative abundance of Symbiodiniaceae OTUs
Now we can calculate the relative abundance of each unique OTU per sample. This will allow us to view the community structure in a faceted barplot.
Caluculate OTU relative abundances
head(its2Norm)## sample Site Depth sq01_C3 sq10_C3 sq11_C3g sq14_C3g sq17_C3g sq18_C3g
## 108 108 TR 10 830134.8 0 0 0.00000 0.00000 0.000
## 109 109 TR 10 828909.5 0 0 0.00000 0.00000 0.000
## 110 110 TR 10 826199.7 0 0 25.04951 12.52476 2580.100
## 111 111 TR 10 826360.2 0 0 0.00000 0.00000 0.000
## 112 112 TR 10 826546.9 0 0 0.00000 0.00000 0.000
## 113 113 TR 10 825665.6 0 0 29.63533 63.50427 3496.969
## sq25_C3g sq05_C3z sq07_C3e sq08_C3g
## 108 0.000000 68607.55 18251.16 0.00000
## 109 0.000000 70578.82 13438.77 0.00000
## 110 4.174918 83289.62 33077.88 150.29706
## 111 0.000000 74603.14 39432.63 0.00000
## 112 0.000000 82377.62 29126.13 0.00000
## 113 59.270653 98321.55 26426.24 88.90598its2NormPerc = its2Norm
its2NormPerc$sum = apply(its2NormPerc[, c(4:length(its2NormPerc[1,]))], 1, function(x) {
sum(x, na.rm = T)
})
its2NormPerc = cbind(its2NormPerc[, c(1:3)], (its2NormPerc[,
c(4:(ncol(its2NormPerc)-1))] / its2NormPerc$sum))
head(its2NormPerc)## sample Site Depth sq01_C3 sq10_C3 sq11_C3g sq14_C3g sq17_C3g
## 108 108 TR 10 0.9052788 0 0 0.000000e+00 0.000000e+00
## 109 109 TR 10 0.9079690 0 0 0.000000e+00 0.000000e+00
## 110 110 TR 10 0.8739716 0 0 2.649790e-05 1.324895e-05
## 111 111 TR 10 0.8787364 0 0 0.000000e+00 0.000000e+00
## 112 112 TR 10 0.8811325 0 0 0.000000e+00 0.000000e+00
## 113 113 TR 10 0.8653400 0 0 3.105935e-05 6.655574e-05
## sq18_C3g sq25_C3g sq05_C3z sq07_C3e sq08_C3g
## 108 0.000000000 0.000000e+00 0.07481792 0.01990326 0.000000e+00
## 109 0.000000000 0.000000e+00 0.07731046 0.01472053 0.000000e+00
## 110 0.002729284 4.416317e-06 0.08810553 0.03499048 1.589874e-04
## 111 0.000000000 0.000000e+00 0.07933163 0.04193194 0.000000e+00
## 112 0.000000000 0.000000e+00 0.08781788 0.03104964 0.000000e+00
## 113 0.003665003 6.211869e-05 0.10304603 0.02769606 9.317804e-05Now a quick sanity check. If this worked the sum of each row should = 100% (i.e. “1”)
# test that all are now 100% = 1
apply(its2NormPerc[, c(4:(ncol(its2NormPerc)))], 1, function(x) {
sum(x, na.rm = T)
})## 108 109 110 111 112 113 114 116 117 118 119 120 121 156 93 94 95 96
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 97 98 99 100 101 102 103 104 105 106 137 122 123 124 125 126 127 128
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 129 130 131 132 133 134 135 136 226 227 228 229 230 231 232 233 234 236
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 237 238 239 240 29 30 31 32 33 34 35 36 37 38 39 40 41 42
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 43 17 18 19 20 21 22 23 24 25 26 27 28 45 46 1 2 3
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 4 5 6 7 8 9 10 11 12 13 14 15 16 211 213 214 215 216
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 217 218 219 221 222 223 224 225 76 77 78 79 80 81 82 83 84 85
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 86 87 88 89 90 91 47 48 49 50 51 52 53 54 55 56 57 58
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 59 60 92 61 62 63 64 65 66 68 69 70 71 72 73 74 75 198
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 199 200 201 202 203 204 205 206 207 208 209 210 241 242 115 152 153 154
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 155 157 158 159 160 161 162 163 164 165 166 148 149 150 151 176 177 178
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 179 180 181 182 183 187 188 145 146 147 167 168 169 170 171 172 173 174
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
## 175 184 185 138 139 140 141 142 143 144 190 191 192 193 194 195 196
## 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1Everything adds up to 1, this is good! The code works.
Add plot order to data frame
I added an additional column to sort better for the stacked barplot. This was just a work around to get the facet_grid() function to play nice with our data. I added a coulumn “barPlotOrder” and for each population I filled in a series 1:n foreach sample at each Site:Depth combo, so now there’s no large blank expanses on the plot.
its2ra = its2NormPerc
sampleCounts = plyr::count(its2ra, c('Site','Depth'))
meltedList = melt(lapply(sampleCounts$freq,function(x){c(1:x)}))
its2ra$barPlotOrder = meltedList$value
colnames(its2ra)[c(5:ncol(its2ra)-1)] = c("sq01_C3", "sq10_C3", "sq11_C3g","sq14_C3g", "sq17_C3g",
"sq18_C3g", "sq25_C3g", "sq05_C3z", "sq07_C3e",
"sq08_C3g")
its2ra=its2ra[c(1,ncol(its2ra),2:(ncol(its2ra)-1))]
head(its2ra)## sample barPlotOrder Site Depth sq01_C3 sq10_C3 sq11_C3g sq14_C3g
## 108 108 1 TR 10 0.9052788 0 0 0.000000e+00
## 109 109 2 TR 10 0.9079690 0 0 0.000000e+00
## 110 110 3 TR 10 0.8739716 0 0 2.649790e-05
## 111 111 4 TR 10 0.8787364 0 0 0.000000e+00
## 112 112 5 TR 10 0.8811325 0 0 0.000000e+00
## 113 113 6 TR 10 0.8653400 0 0 3.105935e-05
## sq17_C3g sq18_C3g sq25_C3g sq05_C3z sq07_C3e
## 108 0.000000e+00 0.000000000 0.000000e+00 0.07481792 0.01990326
## 109 0.000000e+00 0.000000000 0.000000e+00 0.07731046 0.01472053
## 110 1.324895e-05 0.002729284 4.416317e-06 0.08810553 0.03499048
## 111 0.000000e+00 0.000000000 0.000000e+00 0.07933163 0.04193194
## 112 0.000000e+00 0.000000000 0.000000e+00 0.08781788 0.03104964
## 113 6.655574e-05 0.003665003 6.211869e-05 0.10304603 0.02769606
## sq08_C3g
## 108 0.000000e+00
## 109 0.000000e+00
## 110 1.589874e-04
## 111 0.000000e+00
## 112 0.000000e+00
## 113 9.317804e-05Create OTU stack for plotting
Stack the OTU data and adjust columns and factor names for plotting
gss = otuStack(its2ra, count.columns = c(5:length(its2ra[1, ])),
condition.columns = c(1:4))[1:2330,] # remove summ rows
levels(gss$otu)## [1] "summ" "sq01_C3" "sq10_C3" "sq11_C3g" "sq14_C3g" "sq17_C3g"
## [7] "sq18_C3g" "sq25_C3g" "sq05_C3z" "sq07_C3e" "sq08_C3g"gss$otu = factor(gss$otu, levels(gss$otu)[c(1:8, 11, 9, 10)])
levels(gss$otu)## [1] "summ" "sq01_C3" "sq10_C3" "sq11_C3g" "sq14_C3g" "sq17_C3g"
## [7] "sq18_C3g" "sq25_C3g" "sq08_C3g" "sq05_C3z" "sq07_C3e"levels(gss$Depth)## [1] "10" "16" "25" "35"levels(gss$Depth) = c("10 m", "16 m", "25 m", "35 m")
levels(gss$Site)## [1] "TR" "RW" "SR" "GR"levels(gss$Site) = c("Tobacco Reef", "Raph's Wall", "South Reef", "Glover's Reef")
levels(gss$Depth)## [1] "10 m" "16 m" "25 m" "35 m"levels(gss$Site)## [1] "Tobacco Reef" "Raph's Wall" "South Reef" "Glover's Reef"Consruct OTU barplot
OTUplotA = ggplot(gss, aes(x = barPlotOrder, y = count, fill = factor(otu))) +
geom_bar(position = "stack", stat = "identity", color = "black",
size = 0.25) +
ylab("Normalized proportion") +
scale_fill_manual(values=its2ColPal)+
labs(fill = "OTU_Clade type") +
facet_grid(Depth ~ Site, scales = "free_x") + #faceting plots by Depth and Site
theme_bw()
OTUplot = OTUplotA +
theme(axis.title.x = element_blank(),
axis.text.x = element_blank(),
axis.ticks.x = element_blank(),
axis.title.y = element_text(color = "black", size = 12),
axis.text.y = element_text(color = "black", size = 12),
legend.position = "right",
legend.title = element_text(color = "black", size = 12),
legend.text = element_text(color = "black", size = 12),
legend.key = element_blank(),
legend.background = element_blank(),
panel.border = element_blank(),
panel.background = element_rect(fill = "white"),
plot.background = element_blank(),
strip.text.x = element_text(size = 12),
strip.text.y = element_text(size = 12),
strip.background = element_rect(fill = "white", size = 0.9)
)
OTUplotSave the OTU barplot
ggsave("OTUbarPlot2.eps", plot = OTUplot, width = 8, height = 6, unit = "in",
dpi = 600)
ITS2 community differences
Now we can use permutational multivariate analysis of variance (PERMANOVA) to test for differences in Symbiodiniaceae \(\beta\)-diversity across Site and Depth based on our OTUs.
Cheking dispersion
Using betadisper() in vegan to look at multivariate homogeneity of dispersion between sites and depths. This is using Bray-Curtis similarity.
set.seed(694)
anova(betadisper(its2Dist, its2Norm$Depth))## Analysis of Variance Table
##
## Response: Distances
## Df Sum Sq Mean Sq F value Pr(>F)
## Groups 3 0.5242 0.174742 5.0941 0.001971 **
## Residuals 229 7.8553 0.034303
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Dispersion is heteroschedastic, but PERMANOVA is robust to deviations in homgeneity of variance ( Anderson and Walsh, 2013; https://esajournals.onlinelibrary.wiley.com/doi/10.1890/12-2010.1){target=“_blank“}.
Running PERMANOVA in R
Now let’s see how different communities are from each other with PERMANOVA. We will utilize the adonis() function in vegan. We will use Bray-Curtis similarity for our distance matrix and run a total 0f 9,999 permutations, and test the effects of Site, Depth, and the interaction between Site and Depth.
set.seed(694)
its2Adonis = adonis(its2Norm[, c(4:ncol(its2Norm))] ~ Depth*Site,
data = its2Norm, permutations = 9999, method = "bray")
its2Adonis##
## Call:
## adonis(formula = its2Norm[, c(4:ncol(its2Norm))] ~ Depth * Site, data = its2Norm, permutations = 9999, method = "bray")
##
## Permutation: free
## Number of permutations: 9999
##
## Terms added sequentially (first to last)
##
## Df SumsOfSqs MeanSqs F.Model R2 Pr(>F)
## Depth 3 0.3946 0.131523 3.4741 0.04443 0.0063 **
## Site 3 0.0349 0.011633 0.3073 0.00393 0.9065
## Depth:Site 9 0.2364 0.026268 0.6938 0.02662 0.7938
## Residuals 217 8.2153 0.037859 0.92502
## Total 232 8.8812 1.00000
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1We see that Depth has a signifcant effect on Symbiodiniaceae community structure in our M. cavernosa samples.
Pairwise PERMANOVA for multiple comparisons
Since we found that Depth was a significant factor in our PERMANOVA we can now use pairwise PERMANOVA to reveal where differences occur across depth. This utilizes the package pairwiseAdonis, where we will again use Bray-Curtis similarity and 9,999 permutations. We also have added false discovery rate (FDR) corrections since we are perfoming multiple comparisons.
set.seed(694)
its2PWAdonis = pairwise.adonis(its2Norm[, c(4:ncol(its2Norm))],
factors = its2Norm$Depth, sim.method = "bray",
p.adjust.m = "BH", perm = 9999)
its2PWAdonis## pairs Df SumsOfSqs F.Model R2 p.value p.adjusted sig
## 1 10 vs 16 1 3.791992e-02 0.5212379 0.004473330 0.5203 0.62436
## 2 10 vs 25 1 1.215016e-01 4.0807976 0.033703094 0.0003 0.00180 *
## 3 10 vs 35 1 1.209040e-01 3.9840789 0.033767936 0.0029 0.00620 *
## 4 16 vs 25 1 2.553940e-01 5.8424234 0.048347453 0.0031 0.00620 *
## 5 16 vs 35 1 2.551471e-01 5.7118059 0.048523645 0.0108 0.01620 .
## 6 25 vs 35 1 7.599081e-05 0.1802355 0.001592465 0.8301 0.83010We see that again see differences between our deeper (25 + 35 m) and shallower (10 + 16 m) samples.
PERMANOVA without 35 m samples
Subset our dataframe
First we need to remove the deep samples from the dataframe. We will use our dataframe of good OTUs, which haven’t been normalized yet. This way we can calculate the normalization based on only the samples we are keeping in the analysis.
goods2 = subset(goods, !Depth=="35")
goods2[] = lapply(goods2, function(x) if(is.factor(x)) factor(x) else x)
summary(goods2)## newOrder sample Site Depth sq1
## Min. : 1.0 Min. : 1.00 TR:44 10:60 Min. : 224
## 1st Qu.: 59.0 1st Qu.: 46.00 RW:45 16:58 1st Qu.: 90703
## Median :103.0 Median : 91.00 SR:45 25:59 Median :132450
## Mean :109.6 Mean : 92.51 GR:43 Mean :132460
## 3rd Qu.:160.0 3rd Qu.:136.00 3rd Qu.:176321
## Max. :219.0 Max. :188.00 Max. :430489
## sq10 sq11 sq14 sq17
## Min. : 0 Min. : 0 Min. : 0 Min. : 0.0
## 1st Qu.: 0 1st Qu.: 0 1st Qu.: 0 1st Qu.: 0.0
## Median : 0 Median : 0 Median : 0 Median : 0.0
## Mean : 1739 Mean : 1592 Mean : 504 Mean : 363.1
## 3rd Qu.: 0 3rd Qu.: 0 3rd Qu.: 0 3rd Qu.: 0.0
## Max. :53422 Max. :117162 Max. :28206 Max. :13521.0
## sq18 sq25 sq5 sq7
## Min. : 0.0 Min. : 0 Min. : 935 Min. : 0
## 1st Qu.: 0.0 1st Qu.: 0 1st Qu.: 7391 1st Qu.: 2389
## Median : 0.0 Median : 0 Median :11115 Median : 4938
## Mean : 370.1 Mean : 229 Mean :11759 Mean : 5134
## 3rd Qu.: 0.0 3rd Qu.: 0 3rd Qu.:14646 3rd Qu.: 7149
## Max. :9958.0 Max. :8747 Max. :56300 Max. :15419
## sq8
## Min. : 0
## 1st Qu.: 0
## Median : 0
## Mean : 3870
## 3rd Qu.: 0
## Max. :169042Normalize reads
We will normalize samples again, as we did with the entire dataframe orignially
itsgoods2Transposed = t(goods2[, 5:length(goods2[1, ])])
itsgoods2List = DGEList(counts = itsgoods2Transposed)
head(itsgoods2List$samples)## group lib.size norm.factors
## 108 1 237951 1
## 109 1 18070 1
## 110 1 226433 1
## 111 1 204975 1
## 112 1 27762 1
## 113 1 225375 1its2Norm2 = calcNormFactors(itsgoods2List, method = "TMM")
head(its2Norm2$samples)## group lib.size norm.factors
## 108 1 237951 1.080109
## 109 1 18070 1.084610
## 110 1 226433 1.047130
## 111 1 204975 1.052500
## 112 1 27762 1.055361
## 113 1 225375 1.037930its2TMM = t(cpm(its2Norm2, Normalized.lib.sizes = TRUE))
its2Norm2 = cbind(goods2[,c(2:4)], its2TMM)
head(its2Norm2)## sample Site Depth sq1 sq10 sq11 sq14 sq17 sq18
## 108 108 TR 10 838136.9 0 0 0.00000 0.00000 0.000
## 109 109 TR 10 837139.0 0 0 0.00000 0.00000 0.000
## 110 110 TR 10 834635.2 0 0 25.30527 12.65263 2606.443
## 111 111 TR 10 834904.0 0 0 0.00000 0.00000 0.000
## 112 112 TR 10 834911.1 0 0 0.00000 0.00000 0.000
## 113 113 TR 10 833716.8 0 0 29.92431 64.12352 3531.068
## sq25 sq5 sq7 sq8
## 108 0.000000 69268.90 18427.09 0.00000
## 109 0.000000 71279.53 13572.19 0.00000
## 110 4.217545 84140.02 33415.61 151.83161
## 111 0.000000 75374.47 39840.33 0.00000
## 112 0.000000 83211.23 29420.87 0.00000
## 113 59.848614 99280.30 26683.93 89.77292Run PERMANOVA on subset of data
set.seed(694)
its2Dist2 = vegdist(its2Norm2[, c(4:ncol(its2Norm2))], method = "bray")
anova(betadisper(its2Dist2, its2Norm2$Depth))## Analysis of Variance Table
##
## Response: Distances
## Df Sum Sq Mean Sq F value Pr(>F)
## Groups 2 0.3443 0.172163 3.9313 0.02138 *
## Residuals 174 7.6199 0.043793
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1its2Adonis2 <- adonis(its2Norm2[, c(4:ncol(its2Norm2))] ~ Depth*Site,
data = its2Norm2, permutations = 9999, method = "bray")
its2Adonis2##
## Call:
## adonis(formula = its2Norm2[, c(4:ncol(its2Norm2))] ~ Depth * Site, data = its2Norm2, permutations = 9999, method = "bray")
##
## Permutation: free
## Number of permutations: 9999
##
## Terms added sequentially (first to last)
##
## Df SumsOfSqs MeanSqs F.Model R2 Pr(>F)
## Depth 2 0.2828 0.141396 2.92686 0.03326 0.0402 *
## Site 3 0.0367 0.012230 0.25316 0.00432 0.9348
## Depth:Site 6 0.2111 0.035186 0.72835 0.02483 0.6874
## Residuals 165 7.9711 0.048310 0.93759
## Total 176 8.5017 1.00000
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1Depth is still has a significant effect on community structure
Pairwise PERMANOVA
Let’ see where differences lie across depth using pairwise permanova again.
set.seed(694)
its2PWAdonis2 = pairwise.adonis(its2Dist2, factors = its2Norm2$Depth,
perm = 9999, p.adjust.m = "BH")
its2PWAdonis2## pairs Df SumsOfSqs F.Model R2 p.value p.adjusted sig
## 1 10 vs 16 1 0.03164863 0.4486379 0.003852667 0.5956 0.5956
## 2 10 vs 25 1 0.13184809 4.5463670 0.037404384 0.0002 0.0006 **
## 3 16 vs 25 1 0.26179196 6.1925677 0.051096926 0.0016 0.0024 *The differences are still between 25 m and the shallower sites (10 + 16 m). This gives us confidence that our deeper samples aren’t changing our interpretation.
Generalized linear mixed model of OTUs
To see how significantly changing OTUs across depth we will use the package MCMC.OTU, which performs Bayesian analysis of multivariate count data. This will create a MCMC-based GLMM analysis of our OTU data.
Stack OTU data table
its2Glm = its2Norm
colnames(its2Glm)[4:ncol(its2Glm)] = c("sq01_C3", "sq10_C3", "sq11_C3g",
"sq14_C3g", "sq17_C3g", "sq18_C3g",
"sq25_C3g", "sq05_C3z", "sq07_C3e",
"sq08_C3g")
glmStack=otuStack(data=its2Glm, count.columns = c(4:ncol(its2Glm)),
condition.columns=c(1:3))
levels(glmStack$otu)## [1] "summ" "sq01_C3" "sq10_C3" "sq11_C3g" "sq14_C3g" "sq17_C3g"
## [7] "sq18_C3g" "sq25_C3g" "sq05_C3z" "sq07_C3e" "sq08_C3g"glmStack$otu = factor(glmStack$otu, levels(glmStack$otu)[c(1:8, 11, 9, 10)])
levels(glmStack$otu)## [1] "summ" "sq01_C3" "sq10_C3" "sq11_C3g" "sq14_C3g" "sq17_C3g"
## [7] "sq18_C3g" "sq25_C3g" "sq08_C3g" "sq05_C3z" "sq07_C3e"glmStack$count = round(glmStack$count, 0)
head(glmStack)## count otu sample Site Depth
## 1 830135 sq01_C3 108 TR 10
## 2 828910 sq01_C3 109 TR 10
## 3 826200 sq01_C3 110 TR 10
## 4 826360 sq01_C3 111 TR 10
## 5 826547 sq01_C3 112 TR 10
## 6 825666 sq01_C3 113 TR 10Creating the model
Now that we have our data setup correctly we can create the GLMM
Fitting the model
Here we are using Depth as a fixed effect and running 105 iterations with a 5000 rep burn-in period.
set.seed(694)
its2MmD = mcmc.otu(fixed = "Depth", data = glmStack,
nitt=100000, thin=25, burnin=5000)Calculate effect size and p-values
ssD = OTUsummary(its2MmD, glmStack, summ.plot = FALSE)
ssD = padjustOTU(ssD)
sigsD = signifOTU(ssD)
ss2D = OTUsummary(its2MmD, glmStack, otus = sigsD, whiskers = "sd",
ptype = "mcmc")its2MCMCglmD = ss2D$summary
levels(its2MCMCglmD$otu)## [1] "sq01_C3" "sq07_C3e" "sq08_C3g" "sq10_C3" "sq14_C3g" "sq17_C3g"
## [7] "sq18_C3g" "sq25_C3g"its2MCMCglmD$Depth = as.factor(its2MCMCglmD$Depth)
its2MCMCglmD$otu = factor(its2MCMCglmD$otu, levels = c("sq01_C3", "sq10_C3",
"sq14_C3g", "sq17_C3g", "sq18_C3g", "sq25_C3g",
"sq08_C3g", "sq07_C3e"))
ssD$otuWise[sigsD]## $sq01_C3
## difference
## pvalue 10 16 25 35
## 10 NA -0.064565444 0.1631517 0.167166756
## 16 0.46753444 NA 0.2277172 0.231732200
## 25 0.04107614 0.005706208 NA 0.004015015
## 35 0.04096690 0.005638764 0.9926744 NA
##
## $sq10_C3
## difference
## pvalue 10 16 25 35
## 10 NA 2.14263706 -3.7047641 -4.974395
## 16 0.3622074 NA -5.8474012 -7.117032
## 25 0.2025699 0.04107614 NA -1.269631
## 35 0.1287564 0.02956482 0.7971076 NA
##
## $sq14_C3g
## difference
## pvalue 10 16 25 35
## 10 NA -2.19656402 -6.9624223 -6.8118885
## 16 0.041456155 NA -4.7658583 -4.6153245
## 25 0.001261287 0.02407995 NA 0.1505338
## 35 0.001261287 0.02407995 0.9926744 NA
##
## $sq17_C3g
## difference
## pvalue 10 16 25 35
## 10 NA -1.356385955 -7.4099151 -7.31161966
## 16 0.201112944 NA -6.0535292 -5.95523371
## 25 0.001261287 0.005638764 NA 0.09829547
## 35 0.001261287 0.006495356 0.9926744 NA
##
## $sq18_C3g
## difference
## pvalue 10 16 25 35
## 10 NA -3.15991027 -9.6056316 -9.52546355
## 16 0.026922681 NA -6.4457213 -6.36555328
## 25 0.001261287 0.02407995 NA 0.08016806
## 35 0.001261287 0.02407995 0.9926744 NA
##
## $sq25_C3g
## difference
## pvalue 10 16 25 35
## 10 NA -1.252575808 -7.0120491 -6.97676850
## 16 0.220135701 NA -5.7594733 -5.72419269
## 25 0.001653878 0.008451165 NA 0.03528061
## 35 0.001261287 0.007054679 0.9926744 NA
##
## $sq08_C3g
## difference
## pvalue 10 16 25 35
## 10 NA -1.804612359 -8.8973972 -8.87230794
## 16 0.135350191 NA -7.0927849 -7.06769558
## 25 0.001261287 0.005638764 NA 0.02508927
## 35 0.001261287 0.005638764 0.9926744 NA
##
## $sq07_C3e
## difference
## pvalue 10 16 25 35
## 10 NA 0.33996333 0.6058569 0.59265971
## 16 0.0240799535 NA 0.2658936 0.25269638
## 25 0.0001250418 0.06025049 NA -0.01319719
## 35 0.0001850815 0.09119833 0.9926744 NAPlotting the GLMM
We are now plotting the GLMM using our color pallete we created for the OTUs
pd = position_dodge(0.3)
its2glmPlotA = ggplot(its2MCMCglmD, aes(x = Depth, y = mean, group = otu,
colour =otu))+
geom_errorbar(aes(ymin = mean-sd, ymax = mean+sd),
lwd = 0.6, width = 0.5, position = pd) +
geom_line(aes(group = otu), lwd = 0.6, position = pd)+
geom_point(aes(group = otu), position = pd, size = 2.5)+
scale_color_manual(values = its2ColPal [c(1,2,4:8,10)],
name = "OTU_Clade type") +
xlab("Depth")+
ylab("log10(proportion)") +
theme_bw()
its2glmPlotD = its2glmPlotA +
theme(axis.title.x = element_text(color = "black", size = 12),
axis.text.x = element_text(color = "black", size = 12),
axis.title.y = element_text(color = "black", size = 12),
axis.text.y = element_text(color = "black", size = 12),
legend.position = "right",
legend.title = element_text(color = "black", size = 12),
legend.text = element_text(color = "black", size = 12),
legend.key = element_blank(),
legend.background = element_blank(),
panel.border = element_rect(color = "black", size = 1.2),
panel.background = element_rect(fill = "white"),
plot.background = element_blank()
)
its2glmPlotDSave GLMM plot
ggsave("its2glmDepth.eps", plot=its2glmPlotD, width = 8, height = 6, dpi = 600)
Phylogenetic tree of M. cavernosa samples
Finally, to see if the 10 samples with unique Symbiodiniaceae communities are correlated with microsatellite genotypes we will plot a phylogenetic tree with poppr.
Setup data for plotting
set.seed(694)
cbcGenotypes = read.csv("CBCgenotypes.csv")
CBCgen = df2genind(cbcGenotypes[2:10], ind.names = cbcGenotypes$Sample, sep = ",")
CBCgen## /// GENIND OBJECT /////////
##
## // 242 individuals; 9 loci; 150 alleles; size: 185.9 Kb
##
## // Basic content
## @tab: 242 x 150 matrix of allele counts
## @loc.n.all: number of alleles per locus (range: 6-56)
## @loc.fac: locus factor for the 150 columns of @tab
## @all.names: list of allele names for each locus
## @ploidy: ploidy of each individual (range: 2-2)
## @type: codom
## @call: df2genind(X = cbcGenotypes[2:10], sep = ",", ind.names = cbcGenotypes$Sample)
##
## // Optional content
## - empty -CBCdist = provesti.dist(CBCgen)
head(CBCdist)## [1] 0.5555556 0.5555556 0.6111111 0.8333333 0.6111111 0.7222222Contsrtuct phylogenetic tree
Based on genotypes from Eckert, R. J., Studivan, M. S., and Voss, J. D. (2019). Populations of the coral species Montastraea cavernosa on the Belize Barrier Reef lack vertical connectivity. Sci. Rep. 9, 7200. Starred samples in the data set contain the unique C3g dominant communities.
set.seed(694)
MCAVtree = CBCdist %>%
nj() %>% # calculate neighbor-joining tree
ladderize() # organize branches by clade
mcavTree = plot(MCAVtree) %>%
add.scale.bar(length = 0.05) # add a scale bar showing 5% difference.Save the tree
ggsave("MCav_tree.eps", plot(MCAVtree) %>% add.scale.bar(length = 0.05), height = 30, units = "in", dpi = 600)